At Klavis AI, we're building Open Source MCP Integrations for AI Applications. Today, we're making a significant step forward for our open source community: pre-built Docker images for all our MCP servers with OAuth handled for you.
Simplifying Local Deployment for Everyone
Many developers have faced the challenge of running MCP Servers locally, manually setting up numerous dependencies, retrieving access tokens and low quality.
At Klavis, we've made this process simple with just two commands. Starting today, it's easy to pull our MCP Server Docker image to instantly spin up any of our high-quality MCP Server.
That's why we've containerized our entire MCP server ecosystem. From our github repository Klavis-AI/klavis, local deployment is now as simple as a single Docker command:
docker pull ghcr.io/klavis-ai/github-mcp-server:latest
docker run -p 5000:5000 -e KLAVIS_API_KEY=your_key ghcr.io/klavis-ai/github-mcp-server:latest
No compilation. No dependency hell. No OAuth implementation. Just pull and run.
Automatic Updates, Zero Friction
Docker distribution isn't just about easy deployment—it's about staying current without effort:
- Automated updates - Set up watchtower or similar tools for automatic updates
- Version control - Pin specific versions or always run latest
- Rollback capability - Something broke? Roll back to previous version instantly
- CI/CD integration - Integrate our images directly into your pipelines
This means the community can focus on what matters: building amazing AI applications, finding edge cases, and pushing the boundaries of what's possible.
Getting Started in Seconds
We've made deployment incredibly simple. Here's how to run a GitHub MCP server locally:
Quick Start with Klavis OAuth
# Get your API key from https://www.klavis.ai
docker run -p 5000:5000 \
-e KLAVIS_API_KEY=your_klavis_api_key \
ghcr.io/klavis-ai/github-mcp-server:latest
That's it! The container handles all OAuth complexity for you:
- Automatic token acquisition and refresh
- Secure credential management
- Multi-user authentication flows
No need to implement OAuth yourself, no need to manage tokens—just pull and run.
Getting Your API Key
- Sign up for a free account at klavis.ai
- Navigate to your dashboard
- Generate your API key
- Start testing immediately
Custom Authentication (Optional)
If you prefer using your own tokens or have specific authentication needs:
docker run -p 5000:5000 \
-e AUTH_DATA='{"access_token":"your_token_here"}' \
ghcr.io/klavis-ai/github-mcp-server:latest
This flexibility means you can test with our OAuth service or bring your own credentials—whatever works for your use case.
What We've Containerized
Every MCP server in our ecosystem is now available as a Docker image:
- GitHub - Code repositories and issue management
- Gmail - Email integration
- Google Calendar - Calendar access
- GitLab - Alternative Git platform support
- Vercel - Deployment management
- And 50+ more!
All images are:
- Versioned - Pin specific versions for stability
- Lightweight - Minimal overhead, fast pulls
- Self-contained - No external dependencies
- Regularly updated - Latest features without compilation
Empowering the Community to Lead
By removing every possible friction point, we're enabling the community to do what it does best: innovate faster than any company could alone.
Instant Testing:
- Pull our latest changes in seconds
- Test new features immediately
- Report issues with exact version numbers
Rapid Feedback Loop:
- Reproducible environments = faster bug fixes
- Consistent testing = clearer communication
- Lower barriers = more contributors
Community-Driven Evolution:
- Your use cases guide our development
- Your feedback shapes our priorities
- Your contributions make the platform better for everyone
We're not just open-sourcing code—we're open-sourcing the entire development process. When deployment is this simple, every developer becomes a potential contributor.
One More Thing
As AI agents become more sophisticated, they need access to more tools. But here's the challenge: every AI system has practical limits on how many MCP connections it can manage. Claude might handle 10 tools gracefully, but what happens when you need 50? Or 500?
Beyond the connection limits, AI agents face three critical challenges today:
The Tool Selection Dilemma: Agents are paralyzed by choice. Given hundreds of tools from different MCP servers, they struggle to accurately pick the right one for a specific task and often fail to supply the correct parameters.
The Context Window Constraint: Showing AI agents hundreds of MCP tools with detailed descriptions overwhelms the LLM's context window, leading to high costs, slow performance, and errors.
The Tool Coverage Gap: Due to the above issues, most MCP servers offer a limited toolset (often less than 50) and cannot cover a wide range of use cases.
The Missing Context Gap: Humans speak in shorthand - 'the Q3 report,' 'the design channel.' Without the correct context provided by the MCP servers, AI agents can't connect this shorthand to the specific file or channel, leading to failed tasks and frustrating clarifications.
We're actively working on solutions to these fundamental challenges. Our goal is to break through the current limitations of AI agent architectures and enable AI agents that can truly access any tool they need, when they need it, without architectural constraints.
Join our waitlist to be the first to know when we launch it.
Join Us
Ready to get started? Check out our GitHub repository for deployment instructions, or dive into the documentation to understand the architecture.
Have ideas for new MCP servers? Want to improve our OAuth flows? Found a bug? We're all ears. Open an issue, submit a PR, or just star the repo to show your support.
Together, we're not just building tools—we're building the foundation for the next generation of AI applications.